VOLUME
5
•
2004
ISSUE 3
L I N K I N G M E A S U R E S O F C U S T O M E R S AT I S F A C T I O N , VA L U E
A N D L O YA LT Y T O F I N A N C I A L P E R F O R M A N C E
Conventional managerial wisdom holds
that attending to customer satisfaction,
value, and loyalty makes good business
sense for at least two reasons:
1. Satisfied customers are likely to continue
to buy from and/or continue to do
business with a company, while
dissatisfied customers are likely to take
their business elsewhere.
2. Satisfied customers tell others about their
positive experiences, while dissatisfied
customers tell even more people about
their negative experiences.
In fact, faith in the preceding alleged benefits
has set into motion more than a few
customer focus and satisfaction initiatives
in companies throughout the United States
and around the world.
Still, in business as in spiritual matters, faith
will be put to the test. At some point, no
matter what their stated beliefs or
commitments, senior managers, employees,
and shareholders will demand evidence of
the bottom-line impact of customer
satisfaction, value, and loyalty. They will
demand evidence that investments made
in managing relationships with customers,
as well as those made in relevant process
and quality improvements, actually
contribute to growth in revenues,
profitability, and other financial and market
performance indicators.
This paper describes and illustrates three
basic methods of linking measures of
customer satisfaction, value, and loyalty
to financial and market performance
results. The advantages and limitations
of each method also are discussed. Next,
we discuss several factors related to (a)
data structure, completeness, and
availability, and (b) contextual sources of
influence, that may impact the
appropriateness and/or viability of each
of the three methods in any given
organization or industry. Finally, the paper
concludes with guidelines and
recommendations for using the three
methods, individually and in combination,
to establish the “bottom-line benefits” of
customer satisfaction, value, and loyalty.
COMMON APPROACHES TO
LINKING CUSTOMER
S AT I S FA C T I O N , VA L U E A N D
L O YA LT Y T O F I N A N C I A L A N D
MARKET PERFORMANCE
Basically, there are three methods of linking
measures of customer satisfaction, value, and
loyalty to market and financial performance:
1. Projection
2. Direct linkage
3. A hybrid approach that combines
projection and direct linkage
Each of these approaches, along with its
strengths and limitations, is described below.
The Method of Projection
Many of the models developed for
the purpose of demonstrating the financial
or market impact of customer satisfaction,
value, and loyalty rely on projection (e.g.,
TARP, 1986; Hart, Heskett, and Sasser, 1990;
Rose, 1990; Cannie, 1991; Rust, Zahorik,
and Keiningham, 1994; Reichheld, 1996;
Anton, 1997; Vavra, 1997).
In the method of projection, estimates or
assumptions are made regarding the
financial/market worth of a customer. These
estimates usually are derived from such
sources as historical data on customer
accounts, financial records, and/or market
studies. The method works basically by
combining the above estimates with data on
customer satisfaction, dissatisfaction, and
market behavior intentions in order to project
the financial/market impact of success or
failure in addressing customer needs and
requirements, and/or building customer
loyalty and commitment.
The following example, based on work done
in the insurance industry, illustrates the use
of projection to quantify the potential costs
of customer dissatisfaction.
Assume that a company that insures
physicians against malpractice currently
has 6,000 policyholders, and that the
current average annual value of each policy
is $15,000. Assume also that, based upon
a recent sur vey of the company’s
policyholders:
1. 40 percent of the policyholders
experienced some service problem or
failure during the past 12 months (an
estimated 2,400 of the 6,000 total
customer base).
2. 80 percent of these policyholders
reported the problem to the company,
and 20 percent did not (1,920 and 480
of the 2,400 having experienced a
problem, respectively).
3. 40 percent of those who contacted the
company were very satisfied with the
company’s response (768 of the 1,920).
4. 35 percent of those who contacted the
company were only partially satisfied
with the company’s response (672 of
the 1,920).
5. 25 percent of those who contacted the
company were dissatisfied with the
company’s response (480 of the 1,920).
6. All of the very satisfied policyholders
intend to renew their policies during the
coming year.
7. 5 percent of the partially satisfied
policyholders say they will not renew as
a result of their problem experience (34
of the 672 partially satisfied customers).
8. 40 percent of the dissatisfied
policyholders say they will not renew as
a result of their problem experience (192
of the 480 dissatisfied customers).
9. 10 percent of the policyholders who
experienced a problem, but did not
report the problem to the company, will
not renew as a result of their problem
experience (48 of the 480 who did not
report the problem).
Based on the above data, which are
summarized in Figure 1, it is projected that
a total of 274 customers will not renew
their policies (34 partially satisfied customers
+ 192 dissatisfied customers + 48 customers
not reporting the problem). At an average
annual value of $15,000, the projected
cost of customer dissatisfaction is
$4,110,000 — not counting additional
revenues that may be lost due to negative
word-of-mouth communications from the
disgruntled former policyholders who do
not renew their policies.
Given the limitations discussed above, some
managers are skeptical of and/or hesitant
to use the method of projection. Instead,
these managers seek to establish a direct
link between measures of customer
satisfaction, value, and loyalty and actual
financial or other key business results.
The Method of Direct Linkage
An alternative approach to demonstrating
the bottom line benefits of customer
satisfaction is to establish direct
linkage between customer satisfaction,
value, and/or loyalty indicators, and
measures of actual market or financial
performance (Tornow and Wiley, 1991;
Wiley, 1996; Vavra, 1997; Rucci, Kirn,
and Quinn, 1998; Charles, 1999, Allen
and Wilburn, 2002).
The following example provides an
illustration of the method of direct linkage.
In 1993, the authors conducted a national
survey among business decision makers
who were responsible for selecting computer
hardware and software vendors for their
respective companies. Each decision maker
40% Satisfied
0% Will Not
Repurchase
35% Mollified
5% Will Not
Repurchase
80%
% Customers
Who Complain
=
40%
Customers
Experiencing
Problems
274 Lost
Customers
40% Will Not
Repurchase
25% Dissatisfied
20%
% Customers
Who Do Not
Complain
10% Will Not
Repurchase
Figure 2: Stated Customer Satisfaction/Loyalty and Actual Customer Retention
P E RC E N TAG E O F R E TA I N E D C U STO M E R S
Unfortunately, the method of projection
also suffers from several shortcomings: (1)
Projection is highly dependent on
assumptions regarding the financial or
market worth of a customer, and to the
extent that these assumptions are erroneous,
so will be the financial/market impact
projections; (2) This approach also relies
heavily on measures of behavioral intent,
and it is well known that what buyers say
they are going to do in the future is not
always confirmed by their actual purchases
or consumption behavior; and (3) Unless
an organization makes an effort to track
actual market behaviors or financial results,
it is impossible to confirm or refute
projective analyses, and thus, the goodness
of the projection becomes nearly impossible
to evaluate.
Figure 1: Projecting the Financial Impact of Customer Problem Experiences
90%
88%
80%
70%
57%
60%
50%
40%
30%
30%
25%
20%
10%
0%
SE C UR E D
SAT I SFI E D
V ULNE R A BLE
D I SSAT I SFI E D
C USTOME R LOYA LT Y SE G ME NT
Figure 3: Stated Customer Satisfaction/Loyalty and Change in Share of Business
90%
P E RC E N TAG E O F C USTO M E RS
As with any analytical approach, the
method of projection has its strengths
and limitations. Key advantages of this
method are: (1) It does not require a
tremendous amount of data, and the data
required are relatively easy to obtain; (2)
Projection is relatively straightforward —
it need not employ sophisticated or
complex statistical or other computational
techniques (although these may be
utilized); and (3) Projection has high
heuristic appeal because it provides a
relatively simple means of enabling
managers to visualize the ramifications
of improving customer satisfaction and/or
reducing dissatisfaction.
80%
70%
83%
Increased
74%
Decreased
64%
60%
48%
50%
40%
33%
30%
20%
20%
14%
8%
10%
0%
SE C UR E D
SAT I SFI E D
V ULNE R A BLE
C USTOME R LOYA LT Y SE G ME NT
D I SSAT I SFI E D
was asked to evaluate his/her satisfaction
with and commitment to a specific vendor
with whom his/her company currently was
doing business. Based on responses to
questions regarding overall satisfaction,
likelihood to recommend the vendor, and
desire to continue doing business with that
vendor in the future, each decision
maker was placed into one of four customer
loyalty segments:
1. Secured - decision makers who said they
were very satisfied with, definitely would
recommend, and definitely would
continue doing business with this vendor
2. Satisfied - decision makers who said they
were basically satisfied with, probably
would recommend, and probably would
continue doing business with this vendor
3. Vulnerable - decision makers who said
they were marginally satisfied with, might
or might not recommend, or might or
might not continue doing business with
this vendor
4. Dissatisfied - decision makers who said
they were dissatisfied with, probably/
definitely would not recommend, or
probably/definitely would not continue
doing business with this vendor
In 1994 the same decision makers were
again surveyed to determine if they were
still doing business with the same vendors,
and if so, whether they were doing more,
less, or about the same volume of business
as in 1993.
Results, illustrated in Figure 2, indicate a
clear and positive relationship between
customer satisfaction and loyalty (as measured
in 1993) and actual customer retention (as
measured in 1994). Similarly, results
presented in Figure 3 show that decision
makers in the secured segment were (a) less
likely to have reduced, and (b) more likely
to have increased a vendor’s share of business
volume during 1994, than decision makers
who reported lower levels of customer
satisfaction with and commitment to their
respective vendors one year earlier. In short,
the greater the level of customer satisfaction
and commitment reported by a customer,
the more likely a vendor was to retain and
gain share of that customer’s business in
the future.
Of course, the preceding example does not
represent all of the possible ways to establish
direct linkages among customer and
financial/market metrics, but it illustrates the
essence of this method, and highlights its
main distinction from the method of
projection: Direct linkage focuses on actual,
as opposed to possible and/or probable
financial/market results.
As with the method of projection, direct
linkage has its strengths and limitations. The
key advantages of the method of direct linkage
are: (a) It focuses on actual market behaviors
and outcomes, and as a result, frequently has
stronger “face validity” with senior managers
and other stakeholders who are seeking
evidence of the financial/market impact of
customer satisfaction, value, and loyalty; and
(b) Because it focuses on actual market
behaviors and outcomes, direct linkage
provides a means of developing predictive
models and conducting simulations.
Direct linkage also has some limitations.
One limitation is that data requirements
are more complex than in the case of
projective techniques. The effort required
to locate and integrate both customer and
financial/market data is considerably greater
than is the case with projection (more on
this later). An additional challenge centers
on direct linkages that are based on
longitudinal data. Longitudinal analysis
requires willingness to invest in a customer
satisfaction/market performance database.
This demands significant commitment from
and patience on the part of organizational
management that the method of projection
typically does not.
A Hybrid of Projection and Direct
Linkage: Simulation
It is one thing to be able to determine or
project the degree to which measures of
customer satisfaction, value, and loyalty are
linked to and/or will impact financial and
market performance results. However, what
managers and decision makers also seek is an
answer to the question, “in which areas of the
marketing mix will investments in
change/improvement efforts most likely be
profitable (i.e., where will I get the biggest
bang for the buck)?”
Increasingly, managers and decision makers
are attempting to address this question via
a hybrid approach that combines projection
and direct linkage for the purpose of
performing simulations (Allen and Wilburn,
2002; Johnson and Gustaffson, 2000). The
method of simulation seeks to: (1) quantify
the presence or absence of linkages between
quality and service elements and financial
outcomes, (2) determine the strength and
leveragability of such relationships, and (3)
rank alternative quality and service
improvement strategies for managerial action
based on expected (i.e., projected) financial
return to the enterprise through a cost/benefit
analysis. A key benefit of the simulation
approach is the ability to make a linkage
analysis “come alive” and appeal directly to
the desires of CEOs and CFOs for relevancy.
A well built simulator pushes the sometimes
distracting technicalities of data mining and
statistical analysis into the background while
at the same time providing easy and appealing
access to the numbers cast in a language
that management understands — numbers
and figures related to financial and
market performance.
Two primary activities are involved in building
an ROI-driven simulator: (1) Estimating the
payback potential of improvement activities
and (2) Estimating the investment
requirements of improvement activities. These
two estimates are combined in the simulator
in order to calculate the ROI potential of
alternative product, service, or process
improvement strategies.
Estimating Payback. A common tool for
estimating payback in the context of a
simulator is a multivariate statistical model
of the relationship between customer
perceptions of performance and an outcome
(criterion) variable such as observed
customer retention, customer longevity, or
customer value in dollars. It is imperative
that the model builder work closely with
the intended end users (typically managers
or other decision makers) to specify a
conceptual model prior to actual data
analysis. Such a cooperative effort allows
the model builder to understand how
management believes underlying business
processes impact customer perceptions, and
it allows management to gain a realistic set
of expectations regarding what a decision
support simulator can and cannot
realistically provide.
Acquiring and integrating data can be a very
challenging part of the process because it
often requires the model builder to cut across
“functional silos” within an organization.
Often, owners of “voice of the customer”
and “voice of the process” data use different
forms of software or differing file formats.
Further complications arise when silo members
do not speak the same corporate language,
and in many organizations, do not typically
know each other very well. These potential
barriers speak to the need for support from
high in the organization for simulator building
activities. A champion in upper management
is often required to enable the creation of a
multidisciplinary design team or task force
consisting of representatives from voice of the
customer, voice of the process, and IT areas
with a common purpose and the authority
to pull the pieces together in a timely manner.
Decision support simulators are intended to be
prescriptive management tools, that is, tools that
imply specific action. They do not replace
management decision-making, but enhance it
by focusing attention on the most critical variables
in an action-oriented context. If a researcher is
interested only in the predictive ability of a model
in a non-simulator context, then the fact that
specific predictor coefficients behave erratically
because of interdependence (i.e., multicollinearity)
may not be of great concern, as long as there is
a high degree of confidence in the predictions of
the model. To be truly prescriptive, a decision
support simulator must meet a higher level of
clarity — it must provide an accurate prediction
of market outcome and enable the decision maker
to identify which performance variables drive the
outcome. This is necessary for decision makers
to (a) identify what types of corrective actions
might be needed and (b) frame how to estimate
the cost of improvement for ROI calculations.
Because of the prescriptive requirement of
simulators, great care must be taken in selecting
the appropriate type of statistical model to
power the payback function. Issues to consider
in selecting the most appropriate type of
model include the nature and scaling of both
predictor and criteria variables, assumptions
about the linear or non-linear nature of the
data, and identification of the level of
interdependence among predictor variables.
A complete understanding of the availability,
completeness, and structure of relevant data
and potential contextual sources of influence
on the model are critical as well, and will be
discussed in more depth later in this paper.
Estimating Investment Requirements.
While much previous discussion has focused
on the criterion variable for purposes of
modeling payback, an outcome variable of
particular interest to many decision makers
is return on investment (ROI). For example,
a financial services firm could improve the
average rating of “Customer service reps
always being available when I need them”
by 1 point on a 10 point scale, yielding an
estimated 3% increase in customer retention
and $15,000,000 in incremental annual
revenue. Such a finding may be misleading
without knowing what level of investment
will be required to accomplish the 1 point
increase. This improvement could be
accomplished in a number of ways, each
with differing investment requirements
to the firm.
Since a simulator is only as good as the quality
of its inputs, it is critical to estimate as closely
as possible the investments associated with
candidate improvement strategies. One
approach for determining the internal cost
implications of an improvement initiative is
the creation of an action team by the firm
that is charged with identifying high potential
customer hot buttons, mapping internal
processes to those hot buttons, and specifying
ownership within the firm of the appropriate
processes and responsibility for creating action
plans. This type of action planning process is
often facilitated by consultants and involves
creating a cross-functional internal task force
of mid-level and senior managers. The ultimate
deliverables from this group are
recommendations to senior management for
each critical customer issue in terms of possible
action paths, who should take action, timelines
for action, investment requirements for each
proposed action, and expected ROI.
A key tool typically used by action teams is the
type of ROI-based simulator described in this
paper. The needs of the team and the ROI
simulator are intermeshed. The simulator helps
the team hone in on issues that deserve closer
inspection, and the action team ultimately
feeds investment requirements back to the
simulator. Action teams use the payback
function of the simulator to help develop a
short list of customer issues based initially only
on potential payback (i.e., raw estimates of
increased customer value in the absence of cost
or investment information). Ultimately, the
action team revisits its initial short list of
proposed strategies and refines the priorities
for action based on refined ROI estimates made
possible by the team’s process analysis activities.
The following example provides an illustration
of the method of simulation.
In 2000, the authors conducted a customer
satisfaction/loyalty survey for a large North
American financial institution. The client
firm provided customer level information
indicating types and number of products or
services purchased, duration of relationship,
and annual levels of revenue and profitability
represented by each customer. These data
were merged into a master database
representing both attitudinal and behavioral
measures. Based on responses to questions
regarding overall satisfaction, likelihood to
recommend, and desire to continue doing
business with the firm in the future, each
customer was placed into one of four customer
loyalty segments previously defined (Secured,
Satisfied, Vulnerable, Dissatisfied). Preliminary
analysis indicated that the dollar value of a
customer increased with the level of customer
loyalty. This suggested that if a larger
proportion of the customer base could be
moved from “satisfied” to “secured,” then the
total cumulative value of the firm’s customer
base would increase (i.e., the method of
projection). A multivariate model was fitted
to predict loyalty segment membership as a
function of customer perceptions and was
validated with an extensive holdout sample.
This formed the payback engine portion of
a decision support simulator from which the
dollar payback potential could be calculated
for alternative hypothetical improvements in
selected customer perceptions. Internal client
action teams were then able to estimate
investment requirements in order to enable
ROI calculations. Ultimately, evaluation of
the results of several alternative ROI
calculations enabled managers of this financial
institution to choose specific improvement
strategies based on the relative superiority of
their projected cost-benefit outcomes.
As with the methods of projection and direct
linkage, simulation has its strengths and
limitations. Since it is a hybrid of both projection
and direct linkage, it carries with it all of the
strengths and weaknesses already discussed for
those methods. Additionally, its greatest strength
lies in its ability to speak directly to the financial
performance concerns of the most senior levels
of management, and to be prescriptive and
action oriented. The greatest weakness of the
simulation approach is that simulators can be
misunderstood and/or abused by senior managers
and their staffs who might be tempted to use
them too heavily as decision making tools rather
than decision support tools.
To summarize, projection and direct linkage,
as well as a hybrid approach that combines
these two methods, all have certain strengths
as well as limitations. Each method may be
potentially useful in building a case for the
bottom-line benefit of customer satisfaction.
The key is knowing when and how to use
each method most effectively.
FA C T O R S I N F L U E N C I N G T H E
A P P R O P R I AT E N E S S A N D
VIABILITY OF THE THREE
M E T H O D S I N A LT E R N AT I V E
INDUSTRIAL AND MARKET
CONTEX TS
How can an organization determine whether
to use the method of projection, the method
of direct linkage, or a hybrid of the two?
A number of factors can influence how the linkage
between measures of customer satisfaction,
value, and/or loyalty, and financial and
marketplace performance indicators, may be
established or demonstrated. Among these are:
1. Availability, completeness, and structure
of relevant data
2. Contextual sources of influence
Each of these factors is described below in
greater detail.
Availability, Completeness, and
Structure of Relevant Data
Whether an organization can or should apply
the method of projection, the method of direct
linkage, and/or a hybrid of the two, is dictated
in part by the availability, completeness, and
structure of relevant customer and financial/
market performance data.
Where and with whom relevant data reside
plays a major role in determining the best
approach to linkage analysis (and, for that
matter, whether linkage analysis is possible at
all). Within an organization, who owns the
different types of data? How are these owners
accustomed to analyzing, reviewing, and/or
using these data? What, if any, sources of
resistance may create obstacles to acquiring
and assembling all of the data required to
perform linkage analyses? Even if relevant
customer and financial data exist, there is no
guarantee that these data can easily be obtained
and used to populate a template for linkage
analysis. As Schiemann (1996) has observed:
“Organizations face some significant problems
when they must integrate information from
(different data sources) in order to achieve
alignment...Data from these different sources
historically have been guarded by different functions,
such as finance, human resources and marketing.
These guardians have been reluctant to share
information that is part of their turf. After all,
information has given them power and has been
their unique way of adding value to the
organization. Sharing information opens them to
a variety of risks — loss of ownership, control, and
power — that they often would prefer to avoid.”
The completeness or absence of relevant data
also plays a key role in determining what/how
linkage analysis is performed. Complete data
means that the value of each sample unit is
observed or known. For example, if the entire
purchase history of a customer is known and
recorded, then such a string of observations
might constitute complete behavioral data for
the customer. In practice, however, such data
are almost never available, and the analyst
instead must work with censored data, of which
three types have been discussed in some detail
(Wheat and Morrison, 1990):
1. Right-censored data
2. Interval-censored data
3. Left-censored data
The most common case of censoring is what
is referred to as right-censored data. For
example, if some customers did not have a
chance or occasion to repeat-purchase the
product of interest since their current or recent
satisfaction/loyalty measures were gathered,
it would not be possible (at least in the nearterm) to establish the attitudinal-behavioral
linkage of interest. The term right censored
implies that the event of interest is “to the
right” of our currently available data point.
If we continue to monitor those customers’
behaviors, eventually, a relevant repeatpurchase occasion may occur. However, if
the purchase cycle is a relatively long one (for
example, the purchase of a home), the wait
may be considerable.1
Data structure (i.e., form and format) also
plays an important role in determining
what/how linkage analyses may be
performed. For example, do all relevant data
reside on the same or different platforms?
How are they organized?
Generally speaking, data can be characterized
and distinguished as being:
1. Cross-sectional or longitudinal
2. At the individual or aggregate level
of analysis
Some analyses attempt to establish a link
between customer satisfaction, value, and
loyalty, and business results, based on data
collected at a single point-in-time (or
timeframe). Such data are characterized as
cross-sectional. By way of contrast,
longitudinal data consist of observations
or measures collected over multiple pointsin-time. This distinction has several
important implications for linkage analysis.
For example, the statistical tools that are
appropriate for cross-sectional data (e.g.,
OLS regression) may be inappropriate for
longitudinal data, for which other
techniques may be better suited (e.g.,
repeated-measures ANOVA, ARIMA timeseries modeling).
When the individual customer is treated as
the unit of analysis, the focus is on linking
that individual’s satisfaction, loyalty, etc. to
his/her past, recent, and/or future buying
behavior. Each individual has a numerically
coded value for both the customer
satisfaction/ loyalty measure and the
business results indicator. In contrast,
aggregate models utilize group-level
measures (e.g., customer segments, business
units, branches and outlets, aggregate scores
at multiple points-in-time, etc.) as the units
of analysis for both satisfaction/loyalty and
business results.
The level or unit of analysis, in part,
dictates what analytical techniques may
or may not be appropriate, and ultimately,
what/how linkage analyses are performed.
Also, whether individual or aggregate level
data are analyzed directly impacts
interpretation of and inferences made from
results. For example, if aggregate level data
are utilized, great care must be taken when
attempting to use the results to make
inferences about individual customers in
order to avoid the so-called “ecological
fallacy” (Robinson, 1950).
Contextual Sources of Influence
In addition to the availability, completeness,
and form of relevant data, a number of
contextual factors can influence an
organization’s ability to link measures of
customer satisfaction, value, and/or loyalty,
and financial and marketplace performance
indicators. Among these are:
1. Competition and availability of buyer options
2. Degree of market differentiation
3. Business or market cycles
4. Consumption cycles
5. Formality of the buyer-seller relationship
6. Progressive effect of customer expectations
and experiences
Each of these factors is described below in
greater detail.
Competition and Availability of
Buyer Options
Jones and Sasser (1995) suggest that the
relationship between customer satisfaction
and repeat business may vary according
to the competitive environment and
availability of buyer options in a given
industry. These authors identify and
distinguish four types of customers based
on such variations:
1. Loyalists and Apostles - These are
customers who continue to do a relatively
high volume of business with a company
because they are satisfied with that
company, its products, and its services.
Within the loyalist camp are individuals
who are so satisfied, whose experiences so
far exceed their expectations, that they
share their strong feelings with others.
They are apostles.
2. Defectors and Terrorists - These are
customers who, because of their
dissatisfaction, choose not to do repeat
business with a company. The most
dangerous defectors are terrorists. These
are customers who have had a bad
experience, and who can’t wait to get
back at the company by telling others
about their anger and frustration.
3. Mercenaries - These are customers who
are satisfied with a company, but who
demonstrate little or no loyalty in the form
of repeat business. These customers often
are expensive to acquire and quick to
depart. They generally do not remain in
the business relationship long enough to
return a profit.
4. Hostages - These are customers who
continue to do a relatively high volume
of repeat business with a company, despite
their dissatisfaction with that company,
because they have no alternatives and are
basically stuck. However, hostages can
make a company’s life miserable through
constant complaining and negative
word-of-mouth.
Of the four types of customers, mercenaries
and hostages clearly are the most challenging
The point is that in such monopolistic market
settings, it often is difficult to establish a direct,
positive, and/or clear relationship between
customer satisfaction and buying behavior,
making the method of projection the only
viable approach to linkage analysis.
Degree of Market Differentiation
To the extent that buyers perceive little or
no difference in the quality of products or
services provided by competitors in a
marketplace, such products and services are
essentially viewed as commodities, and price
becomes the dominant basis of competition.
In fact, lack of perceived differences among
product/service alternatives is one of the
main reasons that otherwise satisfied
customers become “mercenaries.” Under
such circumstances, an organization may
have considerable difficulty in linking
measures of customer satisfaction or
customer-perceived quality to financial or
market results.
Gale (1994) describes how Ray Kordupleski
and his colleagues at AT&T struggled
during the 1980’s to establish a link between
changes in the company’s index of customer
satisfaction, and changes in customer
retention, defection, and market share.
Although AT&T consistently observed
CUSTOMERS
80%
70%
60%
71%
Satisfied
59%
Dissatisfied
50%
OF
Consider results from a recent study of
customer satisfaction with cable television
service, presented in Figure 4, in which an
attempt to establish a direct link between
customer satisfaction and buying behavior
was made. Compare the behavioral
responses of Dissatisfied cable customers
to those of Satisfied ones. During the sixmonth period after they were surveyed,
significantly more dissatisfied customers
either discontinued or downgraded their
use of cable service than did satisfied
customers, which one would expect if
customer satisfaction and buying behavior
are positively related. However, note that
a slightly higher percentage of dissatisfied
customers actually upgraded their service
than did satisfied ones. In the absence of
an alternative, these disgruntled customers
may have decided to bite the bullet and
spend more with the cable company despite
their dissatisfaction.
Figure 4: Customer Satisfaction and Changes in Level of Service
40%
P E RC E N TAG E
(and often, troublesome) when it comes to
demonstrating a direct link between
customer satisfaction and market behavior
or financial results. In situations in which
one provider has a monopoly, such behaviors
as brand switching, shifting share of
purchases, etc. usually are not possible.
Alternatively, the behavioral response of a
hostage to dissatisfactory service may not
follow a logical or consistent pattern.
30%
20%
10% 12%
10%
0%
D IS CO NNE CT
15%
12%
7%
D OW NG RA D E
M A INTA IN
15%
UPG RA D E
CHANGE IN SERVICE LEVEL
customer satisfaction levels of 90 percent
or higher, the company was losing
customers during the same time period.
Only after it was discovered that many
buyers perceived they could get satisfactory
or even equivalent service quality from
AT&T’s competitors —at comparable or
even lower prices — did the company shift
its focus from an index of customer
satisfaction to a measure of customer value.
Specifically, when AT&T turned its
attention to the customer’s perception of
whether products/services received were
“worth what you paid for them,” the
company was able to create a value index.
Soon after Kordupleski and his colleagues
began tracking this value index, they
discovered that, with a few months’ lag,
it was strongly correlated with changes in
market share.
Levitt (1983) asserts that “there is no such
thing as a commodity. All goods and services
can be differentiated and usually are.”
Nevertheless, when buyers perceive little
difference in the quality of goods and services
offered by alternative brands or firms, there
is a good chance that these buyers will shift
their basis of differentiation to price. Under
such circumstances, measures of customer
satisfaction and/or customer-perceived quality
may have little or no direct correlation with
buyer behavior, while alternative measures,
such as customer-perceived value, may be
more clearly linked to such behavior.
The point of the preceding illustration is
that management must carefully examine
the degree to which buyers differentiate
among alternative brands or firms — as
well as the basis for such differentiation —
in order to select the most appropriate
customer and/or market performance
metrics for linkage analysis.
Business or Market Cycles
In some industries, buyer demand
periodically outstrips the available supply of
a given product and/or the ability of the
producers to provide it. Under such soldout market conditions, it may be difficult
to establish a direct link between customer
loyalty or preference, and actual buying
behavior. For example, buyers of industrial
chemicals may well have a preferred supplier
for such products. However, if the preferred
supplier is not able to provide this product
in the quantities required by the customer
(or at all), these customers probably will
attempt to buy the product from an
alternative supplier. In this case, the dynamics
of product demand and availability take over,
and the positive effects of customer
satisfaction, preference, and loyalty become
nearly impossible to detect — even though
they might be quite evident under ordinary
market conditions.
Thus, when attempting to link measures of
customer satisfaction, value, and loyalty to
business results, organizations must be acutely
aware of the presence of such market
conditions at any given time in their industry.
Failure to do so could lead to the erroneous
conclusion that there is no connection
between customer satisfaction/preference
and buying behavior. Such a conclusion, in
turn, might discourage management from
making necessary investments in building
customer satisfaction and loyalty. An
unfortunate (and ironic) consequence of
such a managerial decision would be the
eventual decline in business results.
Consumption Cycles
We alluded to the potential problems created
by purchase and consumption cycles in
connection with the earlier discussion of
censored data. However, the potential impact
of such cycles is worth a further look.
For products and services that are purchased
and consumed on a regular and relatively
frequent basis, it is not difficult to adopt an
approach in which measures of the customer's
satisfaction, perceived value, likelihood to buy
again, etc. are linked to his/her subsequent
purchases. In fact, in some industries, it may
take relatively little time to establish a direct
link between customer satisfaction and
business results using such an approach.
In contrast, for products and services that
have a relatively infrequent or long-term
purchase and consumption cycle, such an
approach may not be viable. If a real estate
company wishes to determine the extent
to which seller satisfaction leads to repeat
business, that company must confront an
interesting challenge: Given that ten or
more years may pass before the seller once
again has need of the company’s services,
adopting an approach in which each seller's
behavior is traced to his/her satisfaction
with the previous sale is neither practical
nor efficient. Instead, the real estate
company might opt to examine the
relationship between seller satisfaction and
sales volume across different geographical
areas or markets (if the company’s
organizational structure makes this possible).
Alternatively, the company could begin
monitoring and correlating trends in seller
satisfaction with changes in some measure
of business performance — such as the
incidence of repeat customers, or the
percentage of clients having selected the
company on the basis of seller referrals.
whose contracts are up for renewal during
the same points in time.
Progressive Effect of Customer
Expectations and Experiences
United Airlines recently conducted an
exhaustive study of air traveler expectations
and satisfaction. The study took nearly nine
months to complete and included both
qualitative and quantitative research
conducted in eight countries.
Among the study’s key findings was an
inverse relationship between frequency of
air travel and satisfaction with airline service.
Specifically, the study found that:
1. The top 9 percent of the industry’s customers
are the least satisfied with airline service.
2. This same top 9 percent generates
47 percent of the industry’s revenues.
In other words, the customers who bring the
industry the lion’s share of its revenues are also
the least satisfied. In this industry, the
relationship between customer satisfaction and
frequency of usage may actually be negative.
The study goes on to suggest at least two
possible explanations for these findings:
1. Heightened expectations on the part of
frequent travelers
2. Increased opportunity for service failures
as a function of these customers’ volume
of air travel
Customers who are bound by contractual
obligations to buy from or to continue
doing business with a firm may or may not
be satisfied with that firm. Under such
circumstances, the customer basically is a
hostage, at least until the contract expires.
Thus, during the period in which customers
are contractually bound, the firm may find
it difficult to establish a consistent, direct
link between measures of customer
satisfaction and business results, for the
same reasons discussed earlier regarding
market competition and the availability of
buyer options.
Basically, both of the preceding explanations
imply a sort of progressive effect of
customer expectations and service
consumption experiences:
1. To the extent that frequent flyers
increasingly recognize their value to an
airline, they may develop heightened
expectations regarding the quality of service
to which they are entitled. In effect, by
raising the bar, these customers make it
progressively more challenging for an
airline to satisfy their needs and requirements.
2. The more a person flies, the greater the chance
she/he will encounter some service failure
(delays, cancellations, etc.). The hypothesized
cumulative effect of multiple service failures
would be increased customer dissatisfaction.
If the formality of the buyer-seller
relationship is a relevant factor, a firm may
opt to: (a) employ the method of
projection, focusing on the potential
financial or market impact of customer
intentions to new their contract (as in the
insurance illustration presented earlier in
the paper); or (b) begin monitoring
customer satisfaction, monthly or quarterly,
and correlate the results with fluctuations
in revenues retained/lost from customers
The preceding explanations regarding the
progressive effect of customer expectations
and/or consumption experiences certainly
seem plausible. Of course, other factors, such
as the structure and competitive dynamics of
the airline industry (e.g., hub domination by
a single carrier, barriers to switching such as
frequent flyer programs, etc.), also may have
a hand in explaining the findings. Regardless,
the results of the United Airlines study illustrate
how frequency of consumption, alone or in
Formality of the Buyer-Seller Relationship
combination with other market characteristics,
may impact the link between customer
satisfaction and business results.
Guidelines and Recommendations
Clearly, there is no one-size-fits-all approach
to linking measures of customer satisfaction,
value, and loyalty to market and financial
performance. As we have attempted to
show, a variety of factors must be considered
in order to determine whether the method
of projection, the method of direct linkage,
or some combination thereof is most
appropriate for a given organization, market,
or industry.
Mindful of the need to account for the factors
discussed above, the following guidelines and
recommendations are offered:
1. The organization’s strategic business objectives
should guide the selection of financial and
market performance metrics. If a company’s
strategy centers on growth and expansion,
then it should focus on growth rate and/or
market share metrics. Alternatively, if the
strategy centers on leveraging the profit
potential of the existing customer base, then
such indicators as average customer net
present value or customer retention/longevity
may be most appropriate.
2. Valuable learning can be obtained from all
three of the linkage approaches discussed in
this paper. The decision to use the method
of projection, the method of direct linkage,
or some combination of both, should be
directed by an evaluation of the relevance
and presence of the influential factors
discussed above.
3. There is no substitute for a solid
database in which measures of customer
dis/satisfaction, value, and loyalty can be
integrated with indicators of financial
and marketplace performance. Availability,
location, and ownership of customer and
market/financial information may pose
a significant challenge to building such
a database, but this challenge must be
addressed if the organization expects to
go very far with linkage analysis.
4. In the short-term, both projective and
cross-sectional direct linkage techniques can
be used to help senior executives and other
key stakeholders visualize the financial/
marketplace impact of customer satisfaction,
value, and loyalty.
5. In the long-term: (a) simulating the costs
and benefits of alternative intervention,
improvement, and/or change strategies; and
(b) linking trends in customer metrics to
trends in key business indicators, probably
create the most favorable odds of sustaining
the interest of and commitment to customer
satisfaction and loyalty on the part of senior
executives and other key stakeholders.
SUMMARY AND CONCLUSION
The capacity to link measures of customer satisfaction, value, and loyalty to financial and market performance indicators should be part of
the architecture of any organization’s performance measurement and management process. Demand continues to grow for evidence of the bottomline impact of customer satisfaction and loyalty initiatives. Guidance and assurances are sought that investments made in managing relationships
with customers, as well as those made in relevant process and quality improvements, can contribute to growth in revenues and profitability.
Only by demonstrating such linkage can a strong business case for investing in customer satisfaction and loyalty be established.
Both the method of projection and the method of direct linkage offer organizations potentially useful techniques for assessing the financial or
market impact of customer satisfaction, value, and loyalty. Increasingly, managers and decision makers are attempting to fuse marketing and
financial metrics via a hybrid approach that combines projection and direct linkage for the purpose of performing simulation. A well-constructed
simulator allows decision makers to simultaneously assess the payback potential and investment implications of alternative product, service or
process improvement strategies. Such simulators can satisfy the needs of management for profitable prescriptive guidance by integrating “voice
of the customer,” “voice of the process,” and “voice of management” information into an ROI-focused decision support tool.
The keys to using these techniques effectively are: (1) selection and integration of metrics that reflect the organization’s critical success
factors and business objectives; and (2) the ability to recognize and account for the relevance of (a) data availability, completeness, and
structure, and (b) market competition, market differentiation, consumption and market cycles, buyer-seller relationship, and the progressive
impact of customer expectations and experiences on business results.
REFERENCES AND FOOTNOTE
Allen, D. R. and Wilburn, M. (2002). Linking Customer
and Employee Satisfaction to the Bottom Line. Milwaukee:
American Society for Quality.
Andrews, Frank M., Klem, Laura, O'Malley, Patrick M.,
Rodgers, Willard L., Welch, Kathleen B., Davidson,
Terrence N. (1998). Selecting Statistical Techniques for
Social Science Data: A Guide for SAS® Users. Cary, NC:
SAS Institute Inc.
Anton, J. 1997. Call Center Management by the Numbers.
West Lafayette: Purdue University Press.
Brandt, D. R. (1998) “An ‘Outside-In’ Approach to
Defining Performance Targets for Measures of Customer
Service and Satisfaction.” in J.A. Edosomwan, (ed.),
Customer Satisfaction Management Frontiers II, Fairfax,
Virginia: Quality University Press; pp. 7.1-7.11.
Box, G. E. P., and Cox, D. R. (1964). “An Analysis of
Transformations.” Journal of the Royal Statistical Society B
26: 211-43.
Gale, B. T. (1994). Managing Customer Value. New York:
Free Press.
Grapentine, Terry. “Managing Multicollinearity.” Marketing
Research (Fall 1997), 11-21.
Hair, Joseph F. Jr., Anderson, Rolph E., Tatham, Ronald
L., and Black, William C. (1998). Multivariate Data
Analysis. Upper Saddle River, NJ: Prentice-Hall, Inc.
Cannie, J.K. 1991. Keeping Customers for Life. New York:
American Management Association.
Heskett, J. L., Sasser, Jr., W. E., and Schlesinger, L. A.
(1997). The Service Profit Chain: How Leading Companies
Link Profit and Growth to Loyalty, Satisfaction, and Value.
New York: Free Press.
Schiemann, W. (1996). “Driving Change Through
Surveys:Aligning Employees, Customers, and Other
Stakeholders,” in A. Kraut (ed.), Organizational Surveys,
San Francisco: Jossey-Bass.
Jones, T. and Sasser, G. 1995. Why Satisfied Customers
Defect. Harvard Business Review (November/December):
88-99.
Technical Assistance Research Programs (TARP). 1986.
Consumer Complaint Handing in America: An Updated
Study. Washington, D.C.: The Office of the Special Advisor
to the President for Consumer Affairs.
Johnson, M. D. and Gustafsson, A. (2000). Improving
Customer Satisfaction, Loyalty, and Profit: An Integrated
Measurement and Management System. San Francisco:
Jossey-Bass.
Kaplan, R. S. and Norton, D. P. (1996). The Balanced Scorecard.
Cambridge, Massachusetts: Harvard Business Press.
Levitt, T. 1983. The Marketing Imagination. New York:
The Free Press.
Maruyama, Geoffrey M. (1998). Basics of Structural
Equation Modeling. Thousand Oaks, CA: SAGE
Publications, Inc.
Mosteller, F., and Tukey, J.W. (1977). Data Analysis and
Regression. Reading, MA: Addison-Wesley.
Ramaswamy, R. (1996). Design and Management of Service
Processes. Reading, MA: Addison-Wesley.
Robinson, W. S. (1950). “Ecological Correlations and the
Behavior of Individuals.” American Sociological Review
(June): 351-357.
Rucci, A., Kirn, S., and Quinn, R. (1998). “The EmployeeCustomer-Profit Chain at Sears.” Harvard Business Review
(January/February): 83-97.
Charles, B. 1999. Linking research and revenue. Quirk’s
Marketing Research Review 13 (October): 14, 70-72.
Rust, R. T., Zahorik, A. J., and Keiningham, T. L. (1994).
Return on Quality: Measuring the Financial Impact of Your
Company’s Quest for Quality. Chicago: Probus.
Hart, C.W.L., Heskett, J.L., and Sasser, Jr., W.E. 1990.
The Profitable Art of Service Recovery. Harvard Business
Review 68 (July-August): 148-156.
Rust, Roland T., Zeithaml, Valarie A., and Lemon,
Katherine N. (2000). Driving Customer Equity. New
York: Free Press.
Heskett, J., Jones, T., Loveman, G., Sasser, E., Schlesinger,
L. (1994). Putting the Service-Profit Chain to Work.
Harvard Business Review (March/April): 164-171.
Reichheld, F.F. 1996. The Loyalty Effect. Boston: Harvard
Business School Press.
Rose, S. 1990. The Coming Revolution in Credit Cards.
Journal of Retail Banking 12 (Summer): 17-19.
Tornow, W.W. and Wiley, J.W. 1991. Service quality and
management practices: A look at employee attitudes,
customer satisfaction, and bottom-line consequences.
Human Resource Planning, 14: 117-128.
United Airlines. 1997. United Airlines Uses Findings to
Develop New Customer-Driven Initiative. United Airlines
Press Release. May 15, 1997.
Vavra, T.G. 1997. Improving Your Measurement of Customer
Satisfaction Milwaukee: ASQC Quality Press.
Wheat, Rita D. and Donald G. Morrison (1990), “Assessing
Purchase Timing Models: Whether or Not is Preferable to
When,” Marketing Science, 9 ( 2): 162-170.
Wiley, J.W. 1996. “Linking Employee Survey Results to
Customer Satisfaction and Business Performance,” In
Organizational Surveys, edited by A. Kraut. San Francisco:
Jossey-Bass.
1
Interval-censored data reflects uncertainty as to the exact
times the event took place within an interval. This type of
data frequently comes from tests or situations where the object
of interest is not constantly monitored. Current customer
satisfaction data levels with the bank for instance could be
function of the most recent, or the two most recent transactions.
If customer satisfaction data are not observed after every
transaction, then the data might be interval censored. Finally,
left-censored data means that the data to the left of the point
in time are missing. For example, we may have information
only on the two most recent purchase transactions of the
customers, with no record of previous transactions. Such data
are left-censored. Overall, censored data are more often than
not observed in customer satisfaction research because of resource
constraints, and other issues such as the inability to survey
respondents frequently.
B U R K E I N CO R P O R AT E D • 8 0 5 C E N T R A L AV E N U E • C I N C I N N AT I , O H 4 5 2 0 2 •
b u r ke. co m